“Omics” technologies and allograft injury

David Rush

100

50

70

80

90

60

40

0 1 2 3 4 5 6 7 8 9 10 11 12

Time (years)

Graft survival by year of transplantation

Kasiske et al, Am J Transplant (2005); 5:1405

Gra

ft s

urv

iva

l (%

)

1999-2002

1995-19981990-1994

1984-1989 Slope: -2.2 ± 7.22 mL/min/1.73m2/year

Slope: -1.4 ± 10.9 mL/min/1.73m2/year

DeKAF (Decline in Kidney Allograft Function) Study Group

Slope: -0.90 ± 0.18 mL/min/1.73m2/year (aging)*

*Rowe et al, J Gerontol (1976); 31:155

Matas, Kasiske, Hunsicker, Gaston, Mannon, Cecka, Gourishankar, Halloran, Rush

DeKAF Study

Prospective/cross-sectional study (~ 5000 patients)

• Participating centres: Minneapolis (Matas, PI; Kasiske), Mayo

Clinic (Cosio, Grande), Iowa (Hunsicker), Alabama (Gaston, Mannon), UCLA (Cecka), Alberta (Gourishankar, Halloran), Manitoba (Rush)

• Renal Biopsies done “for cause” (n ~ 800) – (Mayo Clinic)

• Urine magnetic resonance done in Winnipeg

Deterioration of Kidney Allograft Function

DeKAF: Hypotheses

1) Progressive graft dysfunction is due to ongoing active injury, and is not necessarily the consequence of past events;

2) There are discrete, definable entities responsible for injury, that lead to chronic graft deterioration and late graft loss;

3) These entities can be differentiated by means of clinical, laboratory, and pathologic studies;

4) Accurate diagnosis offers the best hope for the development of interventional trials.

Gourishankar el al, Am J Transplant (2010); 10: 324

Enrollment: Prospective and Cross-sectional cohorts

Depiction of clusters – “cluster clocks”

Legend

Each spoke represents

a Banff score, except

Length of spokes = %

patients with finding

No spoke = Banff 0

…. = Banff 1

---- = Banff 2

= Banff 3

Inflammation in areas of fibrosis (“iatr”; A)

and tubulitis in atrophic tubules (“tatr”; B)

A B

Inflammation in

areas of scar

Has previously

been excluded

from Banff

Original DeKAF clusters (n = 265; now 370)

Cluster 1 – mild

inflammation;

mild fibrosis

Cluster 2 – ai,

at, iatr, tatr; mild

fibrosis

Cluster 3 – ai,

at, iatr, cg;

fibrosis

Cluster 5 – no

ai, at; only iatr,

tatr; fibrosis

PTC in several

clusters

Actuarial Graft Survival

by Cluster and by “iatr” score

Matas et al Am J Tansplant (2010)

Mannon et al Am J Tansplant (2010)

gDNA

acgtacca

aggtaacg

cggtttttcgt

gtatctccctt

20,000–30,000

Genes

mRNA

cDNA Microarrays

> 100,000

mRNAs

Systems Biology Approach:Profiling all Components in a Sample

Protein

Mass Spectroscopy

> 1,000,000

Proteins

1H Magnetic Resonance

Spectroscopy

Metabolism

Low MW Compounds

(<5kD)

Mueller et al, Am J Transplant (2007); 7: 2712-2722

Microarray analysis of rejection using pathogenesis based transcripts (PBT)

CTL-transcripts

IFN-g transcripts

Transporter

transcripts

4.56 4.36

4.15

3.95

3.74

3.54

3.34

3.13

2.93

2.72

2.52

2.31

2.11

1.91

1.70

1.50

1.29

1.09

PPM

The Human Metabolome Database: Biofluid Urine

785 metabolites identified.

444 have 1H-NMR referenced

spectral peaks.

Wishart et al, Nucleic Acids Research; 37: D603-D610, 2009 (updated 2011)

Urine Metabolomics: Methodology

• The large number of data points in MR spectra requires

an informatics approach:

• The strategy for “pattern recognition” has 4 stages:

– 1) Pre-processing : Area normalization, peak alignment

– 2) Feature selection: Identification of maximally discriminating

averaged subregions of the spectra;

– 3) Classifier development: With these subregions,

crossvalidated linear discriminant analysis classifiers are

developed

– 4) Accurate visualization of resultsSomorjai RL et al. Artificial Intelligence Methods and Tools for Systems Biology (Dubitzky W and

Azuaje F, (eds.)), Computational Biology Series, Vol. 5 Springer pp. 67-85 (2004)

Urine spectra from DeKAF patients

• Matching urine spectra/biopsy pairs (u/b) studied to

date are 457:

– 102 u/b from patients with varying degrees of fibrosis but no

inflammation;

– 150 u/b from patients with varying degrees of fibrosis and

severe inflammation in both normal and atrophic

parenchyma;

– 108 u/b from patients with varying degrees of fibrosis and

minor inflammation mostly in atrophic parenchyma;

– 97 u/b from patients with transplant glomerulopathy.

Moreso et al, Am J Transplant (2006), 6:747

Cosio et al, Am J Transplant (2005) 5: 2464

The combination of inflammation

and fibrosis detected on protocol

biopsy identifies grafts at ↑ risk

1. Can Urine MR spectra distinguish between IF with severe

inflammation and IF without inflammation

• One hundred (100) patients with IF plus severe inflammation and

68 patients with IF minus inflammation were used for the training

set; and 50 and 34 independent patients, respectively, were used as

the test set.

• The 3,300 data point data set of the average spectra was analyzed

(100 points at a time) and the best classifier was developed on the

training set. The classifier was validated with the independent test

set.

• Visualization of the data was done using the “class-proximity

plane” graphic.

Conclusions

• Urine magnetic resonance spectroscopy (UMRS) distinguishes IF with severe inflammation from IF without inflammation with ~90% accuracy.

• The extent of inflammation (severe vs. minor) can also be accurately determined by UMRS with ~90% accuracy.

• Similarly, IF without inflammation can be distinguished from minor inflammation by UMRS with ~ 90% accuracy.

• Validation of UMRS signatures for other allograft pathologies – e.g. transplant glomerulopathy – is in progress.

• The non-invasive nature of UMRS will allow for repeat testing to evaluate changes in spectra and their correlation with specific interventions and their outcomes in prospective studies.

Thank you!